Dragonfly Algorithm + Tabu Search Hybrid Cracks Traveling Salesman Problem
New hybrid metaheuristic combines swarm intelligence with memory-based search for optimal routing.
Deep Dive
Researcher Ammar Bouketta proposes a hybrid metaheuristic for the Traveling Salesman Problem (TSP) by combining the Dragonfly Algorithm (DA) for global exploration with Tabu Search (TS) for local refinement. Tested on TSPLIB benchmarks (burma14, att48, ch150), the hybrid improves tour quality over standalone DA and TS. A systematic grid-search procedure studies the influence of hyperparameters, and results show performance remains sensitive to parameter settings and problem size.
Key Points
- Hybrid combines Dragonfly Algorithm (global swarm search) with Tabu Search (local memory-based refinement) using High-Level Relay Hybridization.
- Tested on TSPLIB instances burma14, att48, and ch150; beats standalone DA, TS, GA, ACO, PSO, and Random Search on most benchmarks.
- Hyperparameter sensitivity (swarm size, tabu tenure) is critical; grid search used for tuning, but performance degrades on larger problems.
Why It Matters
This hybrid approach offers a practical method for optimizing logistics and routing, potentially reducing costs in delivery networks.